R

Ru
T

T
t
At
U

steady state flow rate = the inside radius of a tube at any given point = the unstressed radius of a tube = spatial coordinate in the radial direction = the duration of a pulse = the variable time

Greek Letters

S
p

= thickness of the tubing wall

v
p

= absolute viscosity = kinematic viscosity

= density of flowing fluid = an undefined functional relationship

- t n tn-1 = component of fluid velocity parallel to the axis of

$ I

LITERATURE CITED

AU
U

the tube = axial velocity of the fluid averaged over a cross sectional area of the tube
= Un Un-1 = radial component of velocity

- -tube
Xn

= spatial coordinate in the axial direction of the

-

1. Lambert. T. W., Ph.D. thesis, Purdue University, Lafayette, Indiana ( i956): 2. Morgan, G. W., Bull. Math. Biophysics, 14, 19 (1952). 3. Sheppard, R. G., Ph.D. thesis, University of Iowa, Iowa City -( 1967). 4. Young, Thomas, Trans. Roy. Soc. London, 98, 164 (1808).
Manuscript received April 4, 1967; revision received December 6, 1967; paper accepted Ianuary 8, 1968.

- Zn-1

Effect of imperfect Mixing on Autorefrigerated Reactor Stability

W. L. LUYBEN
Lehigh University, Bethlehem, Pennsylvania This paper discusses the effects of imperfect mixing on the stability of autorefrigerated chemical reactors. Mixing is shown to strongly influence stability in these evaporatively-cooled SYStems. A modified van Heerden steady state stability analysis is presented. Computer simulation of a typical system illustrates the occurrence of runaways and local hot spots. Both continuous stirred tank reactors and batch reactors are considered.

Stability of perfectly mixed continuous autorefrigerated chemical reactors was discussed in a previous paper (1) . The effects of removing the heat of reaction by evaporative cooIing, instead of by conventional cooling jackets or coils, was studied by various stability analysis techniques. The decrease in latent heat of vaporization near the critical temperature was shown to be a source of positive feedback. The purpose of this paper is to consider the case where mixing is not complete so that spacial concentration and temperature gradients can develop. This effect, resulting in local hot spots, would be expected to occur in industrial scale reactors, particularly in highly viscous or slurry systems. Mixing is shown to significantly influence the stability of these evaporatively cooled reactors. A modified van Heerden steady state stability analysis and a computer simulation of a typical system are presented. Both continuous stirred tank reactors and batch reactors are discussed.
SYSTEM

pressures for less volatile bY

B and more volatile A are given

Raoult's law is used to calculate liquid and vapor compositions. The process equipment studied is the typical reactorcondenser system shown in Figure 1. The exothermic heat

, c,c
L X

'W 0

Fw+J-fq
November, 1968

A binary mixtuse of reactant A and product B is considered. Product B is the higher boiler, being of higher molecular weight and stoichiometrically requiring R, moles of reactant A:

B (1) Reaction rate is assumed to be first-order in A. Vapor

R,A--+
Page 880

Fig. 1. Imperfectly mixed autorefrigerated reactor.

AlChE Journal

of reaction vaporizes reactor liquid. Vapor passes to the condenser where its heat of condensation is removed. Liquid returns to one end of the reactor by gravity flow or by pump. In the continuous case, a feed stream is added and product withdrawn. In the batch case no material is removed or added. To model imperfect mixing the reactor is assumed to consist of N perfectly mixed stages. Interstage mixing transports material and sensible heat between stages. The parameter, W, is defined as the interstage mixing rate. It is an effective dispersion coefficient or degree of mixing, and would be related to the system geometry and properties and the agitator speed through the Reynolds and Froude Numbers. Holdup in each stage is assumed constant. The condenser is assumed to consist of perfectly mixed stages on both the process and coolant sides. Holdups of vapor and liquid on the process side of the condenser are assumed negligible. Other assumptions are constant physical properties, phase equilibrium in both condenser and reactor, constant and equal heats of vaporization of A and B.
TEMPERATURE DEPENDENCE OF REACTION RATES

1 . 0

+
3 6 1 r 3

0
600 650 700

750

TEMPERATURE ('Rl

Fig. 2. Temperature dependence of reaction rates.

A preliminary discussion of reaction rates in autorefrigerated systems is presented to introduce the steady state stability analysis. In the staged reactor system defined, liquid vaporizes in each stage of the reactor due to the reaction. Since reactant A is more volatile it is the main component v'd porized. But for component A to be transported clown the reactor to subsequent stages by bulk flow and interstage mixing, concentration differences must exist along the reactor. Of course the higher the degree of mixing, the larger W, the smaller the concentration gradients will be. In the limit as W goes to infinity, mixing is perfect and no gradients exist. As more A reacts from stage to stage, the concentration of B increases for two reasons ( a ) consumption of A and production of 3 by reaction and ( b ) vaporization of mostly A to remove heat. The importance of the latter increases as the ratio of the heat of reaction to the heat of vaporization increases. Since B is a higher boiler, the temperature must increase from stage to stage, for a fixed system pressure, as the concentration of B increases. The higher temperature will tend to increase the reaction rate, consuming and vaporizing still more A. Thus an inherent source of instability exists in imperfectly mixed evaporatively cooled systems. Runaways will occur if mixing is not sufficient to supply enough of component A for vaporization or to transport enough colder material to supply sensible heat removal. It is interesting to note that a runaway will not be unbounded in the system being studied. The reaction rate will tend to increase as temperature increases. But the overall reaction rate, R, is a function of both temperature and the concentration of reactant A.
&TI

= k ( T ) ACT)

(3)

The specific reaction rate, k, increases exponentially with temperature : k = a exp (- E / R T ) (4) The concentration of A however decreases with temperature at a fixed pressure as dictated by vapor-liquid equilibrium in this binary system. Therefore the product of these two effects, as shown in Figure 2, will be limited by k at low temperature and by A at high temperature. The reaction rate will of course be zero at the boiling point

temperature of pure B when there is no A. A nonmonotonic curve results with a maximum reaction rate at some temperature. The location and magnitude of the maximum depends on the activation energy of the reaction, system pressure, and the relative vapor pressures of the components and their temperature dependence. In the region to the left of the maximum, increasing temperature increases the reaction rate. In the region to the right of the maximum, increasing temperature decreases the reaction rate. These regions could be considered the wet side (when reaction rate is limited by temperature) and the dry side (where reaction rate is limited by reactant A concentration). Autorefrigerated reactors can then be operated in either regime. Indeed it is possible in the imperfectly mixed case to have various stages operating in both regimes. However it is clear that the control systems must be quite different for these two cases. For operation on the wet side, a typical control system consists of a pressure (or temperature) controller manipulating coolant to the condenser. If pressure increases due to an increase in reaction rate, more coolant would be used to lower reactor temperature and slow down the reaction. On the dry side, reaction rate is concentration limited. A typical control system would consist of a pressure controller manipulating reactant A feed rate, with coolant flow set at some desired rate. If pressure increases due to an increase in reaction rate, less reactant would be fed into the reactor to lower the concentration of A and slow down the reaction. The temperature in this case would increase. Thus the control actions in the wet and dry regions are opposite. Severe control problems would be expected if one attempts to operate either of these systems at temperatures near the maximum in reaction rate curve. If during an upset the reactor goes over the peak, control action would be in the wrong direction. For example if a dry reactor goes over the maximum, the addition of more feed will decrease instead of increase reaction rate. Reaction rate is now temperature limited and more feed of A lowers the temperature. To operate these systems near the maximum point an adaptive identification technique would be required to make appropriate changes in control action. To illustrate some of the above discussion quantitatively, let the overall reaction rate R be given by Equation ( 3 ) and the specific reaction rate k by Equation ( 4 ) . The concentration of A can be found as a function of temperature for a given system pressure P by using Raoult's law

Vol. 14, No. 6

AlChE

TAULE 1. PARAMETERS

USED

IN

NUhlEHlCAL

EXAhfPLE

A,
A C

CP CW

= -

C 1
c 2 c 3 c 4

E
Fa
AHV

AHU

H W
KEQ
K C

R
R C

T*
Two

U
ff

0.5 wt. fraction 500 sq.ft. 0.75 B.t.u. (lb. - OR.)--' 1.0 B.t.11. (lb. - OR.)-' = -4883.39"R. 12.672 = -44883.39"R. 10.369 = 30,000 B.t.u. (Ib. - mole - OR.) = 2,000 lb./hr. 600 B.t.u./lb. = 274 B.t.u./lb. = 2,400 Ib. 240 lb. = 250 (lb./hr.)/lb./sq.in. = 2,000 (lb./hr.)/( O R . ) 2.99 B.t.u. (1b.mole - "R.)-1 0.549 x 10-6 lb./sq.in. (lb./hr.)-2 530"R. 530"R. - 150 B.t.u. (hr. - sq.ft. - OR.)-' 7.08 x 1010 hr.-1

So finally the temperature dependence of the overall reaction rate can be found explicitly as a function of temperature.
&TI =

[ . I exp (- WRT) 1

P exp(- R T / T - ~ ~ )1
Rp

-1

+ ( P - Rdexp(-

RT/T-Q)

(11)

Figure 3 gives curves of R, expressed as a ratio of the reaction rate at 600"R., versus temperature for several values of activation energy, E , and system pressure, P. These curves are based on the numerical example used in the simulation discussed below. Values of parameters are given in Table 1.

CONTINUOUS REACTOR
M a t h e m a t i c a l Model

The equations describing the system shown in Figure 1 are:

or

1. nth Stage in Reactor total continuity:

Fn-1-

Fn - V n = 0

component continuity: Assuming that the vapor pressure curves for A and B have the same temperature dependence C1 = C 3

RT

(7)
energy:

the ratio of vapor pressures will be constant

dT,, HnCp dt = Cp Fn-1

Substituting into Equation (6) gives

Tn-1-

Cp Fn Tn

- (AH,
kinetic:
.

+ C p Tn) Vn + AH, Hn Rn
Tn

+ W C p (Tw-1-2Tn+
Now weight fraction A is related to mole fraction Xn by

Rn

An . I exp (- E / R T n )

A=

X AM A XAMA ( 1 - x A ) MB X A

X A ( 1 - x A ) R,,, (10)

phase equilibrium:

pn
mass transfer:

= PA(Tn)

xn + P B ( T n ) (1 - xn) V n = K E Q (Pn - PR)

5 '

2. Condenser energy water side:

2
U

O 3
w c
N

E'
<
w a
J I

X 0

Tf=
N

a1 >
0

n=l

n "

600

700

800

xc=

n=l

TEMPERATURE

PR)

Fig. 3. Reaction rates.

Page 882

AlChE Journal

November, 1968

phase equilibrium: pc = P A hydraulic:

(Tc) x c

I
P B (Tc)

(1 -x c )

(22)
.6

3. Controller (Temperature or Pressure) FW = FWO- Kc ( Tset- T i )

Fw = Fwo - Kc ( P s e t - P R )

(24)

The total, component, and energy equations for the first stage in the reactor (n = 1) must be modified to include the liquid stream Lc from the condenser.
STEADY STATE STABILITY ANALYSIS Perfectly M i x e d

a
0

. 3

Van Heerden's steady state technique was first applied to a perfectly mixed reactor. This technique does not yield necessary and sufficient information regarding stability but was employed because of its simplicity and the valuable insight into the process it provides. Heat generat' ion curves QG and heat removal curves Q H , for both temperature and pressure control, are given in Figure 4.
QG QR

0
a
W

I -

= AH, H A a exp (= =

E/RT)

(T-To)(CpFo)
( T - To) C p F o

+ V [AH?+

+ Qc

=
C p

( T - T,)]
0 600
700 T E M P E R A T U R E (OR)
Fig. 4b. Steady state analysis: perfect mixing.

For a specified temperature, A is found from Equation (12) with the time derivative set equal to zero. Qc can then be calculated. Q R is found by simultaneous solution of Equations (16) to (24) applied at steady state to find T , and V. The perfectly mixed reactor is seen to have two stable

open loop ( K , = 0) operating levels, one at a high temperature (720"R.) and one a low temperature ( <55OoR.). A controller gain of about 2,000 is required to operate the reactor close to the desired 600"R.level using temperature control. The heat removal curves with pressure control show that a high gain would be required to operate at the BOOOR. level and that a slight temperature upset will cause the reactor to runaway to the high temperature equilibrium point. This was precisely the behavior experienced when pressure control was attempted in the computer simulation. The explanation of this phenomenon is the nonunique relationship between temperature and pressure in this binary system. As temperature increases, pressure would increase if compositions remained constant. But higher temperatures increase the reaction rate and reduce the amount of the more volatile reactant A . Therefore, pressure wi11 increase with temperature over the lower range but will decrease as the reactor becomes richer in component B.
Imperfectly M i x e d

PRESSURE

CONTROL

/ /

600 TEMPERATURE

700 (OR)

Fig.

417.Steady state analysis: perfect mixing.

A modified version of Van Heerden's steady state stability technique can be used to study the effects of interstage mixing W on the stability of imperfectly mixed autorefrigerated reactors. A simple two-stage reactor is studied with the first stage at some fixed temperature TI. A fixed first stage temperature simplifies the analysis and gives some insight into the process. Actually TI cannot be fixed since it is a dependent variable in the problem. Therefore several values of TI are used. Pressure is assumed constant and equal in both stages. Consider the heat generation and heat removal in the second stage at various temperatures Tz. Heat generation curves can be caIculated directly; fixing pressure P and AlChE Journal
Page 883

Vol. 14, No. 6

:I
3

Q Qz/

Lower mixing rates decrease the speed of response and increase the magnitude of the disturbance. The temperature sensitivity of the reactor increases with the number of stages or as the reactor becomes more distributed. Notice that the 4 stage reactor with a 10,000 mixing rate shows a temperature runaway or hot spot in the last stage. Reactant concentration in the last stage, A4, is seen to decrease despite the increase in feed concentration A. Thus the degree of mixing can drastically change the transfer function between inlet and outlet concentrations, both the time constant and the gain.
TWO STAGES
Ai_ _ - - ---_

_-.

- .3
T2

619

w = 5000

580

600

620 640 TEMPERATURE ( O R )

660

680

Fig. 5. Steady state analysis: imperfect mixing.

temperature

T2,

fixes Az and kz, thus

Q G ~=

l-

a
K
W

HzAzkzAHR

(25)

5 610
W
C 609

.4

I -

Heat is removed from the second stage by sensible heat of the material circulating between the two stages
( Q R ~ ) sensible = W C, ( T z - T I )

I 0

(26)

608
55,

and by vaporization of liquid. Assuming the reactor is operating at the 600"R. level so that the vapor is essentially pure A, the vaporization rate will be limited by the net amount of A being transported into the second stage (the amount entering less the amount reacting). Therefore the heat removed by vaporization will be:
( Q R ~vaporization )
=

,
A2

- .3 a

602
601

0
.2
.4

.6
(HOURS)

.8

1.0

.2

TIME Fig. 60. Transient response.

[W (A1 - Az)

- HzkzAz] AH,
(27)

Heat removal will vary with W and with TI, the value of the first stage temperature chosen. Heat generation curves [from Equation ( 2 5 ) ] and total heat removal curves [from the sum of Equation (26) and (27) for the second stage can be plotted versus second stage temperature as shown in Figure 5. Points of intersection are steady state solutions, and Figure 5 indicates multiple steady state solutions are possible. The steady states will tend to be stable if the slope of the heat removal curve is greater than the slope of the heat generation curve. For example, to operate T z at 6OO"R. with TI at 570"R. or with a 30"R. temperature gradient, an interstage mixing rate of about 1,000 is required. To operate T z at 600"R. with T I at 580"R., with a 20"R. temperature gradient, a mixing rate of about 2,500 is required.
SIMULATION

- =/

'0

.4

g a
I 0

Equations (12) to (24) were simulated on the Lehigh University GE 225 digital computer. Values of parameters are given in Table 1. Steady state conditions for 2 and 4 staged reactors and for various mixing rates are given in Table 2. High interstage mixing rates are required to prevent large temperature gradients. This effect increases as the number of stages increases. Transient responses to changes in feed composition A , from 0.5 to 0.7 wt. fraction are given in Figure 6. Variations in last stage temperature ( T z or T 4 ) are seen to be more rapid and larger than first stage temperature TI.

a
w

602

'"I/ c I.?
.3

- - - - -_ _ - _ - - - - .4
.6
(HOURS)

601

.2

.8

1.0

TIME

Fig. 66. Transient response.

Page 884

AlChE Journal

November, 1968

TABLE 2. STEADY STATECONDITIONS FORN STAGES WITH TEMPERATURE CONTROL Interstage Mixing (lb.1 ( hr. 1 N W n
2 5,000 1 2 10,000 1 2 50,000 1 2 4 10,000 1 2 3 4 20,000 1 2 3 4
BATCH REACTOR

degree of mixing and therefore, mixing should be considered in condenser (design.

CONCLUSION

Reactant

Temperature
Tn (OR.)
601 618 601 609 601 603 602 618 639 656 601 607 612 615

Composition
An

Vapor Rate
Vn

(wt.fraction)
0.268 0.168 0.263 0.211 0.247 0.237 0.263 0.170 0.102 0.070 0.256 0.216 0.187 0.173

(1bJhr.)
424 460 387 434 358 384 296 417 300 100 179 232 210 184

Mixing has a pronounced effect on the stability and performance of both continuous and batch autorefrigerated reactors. Hot spots, runaways, and large temperature gradients can occur unless mixing rates are high. These effects become more pronounced as the system becomes more distributed, affecting reactor steady state and dynamic performance. Autorefrigerated reactors can be operated wet or dry with reaction rate increasing or decreasing with temperature. A control problem can be expected at the transition point.
NOTATION

A A,

c,
c1 c z c s

= reactant concentration-wt. fraction = condenser area, sq.ft. = specific heat reactant and product, B.t.u. (1b.OR.) -1

Mixing can have a pronounced effect on the stability of batch autorefrigerated ,reactors. Figure 7 gives results of a simulation of a batch reactor for various values of W. The same equations were used but with no feed or product withdrawal. A four stage reactor was considered with all stages initially at 580"R. and A = 0.5. Temperature controller setpoint was ramped from 580 to 680"R. in 1 hr. High mixing rates are required to prevent large temperature gradients. The last temperature stage, T4, goes through a maximum before being cooled back down. Notice that the condenser heat duty is affected by the
I

, .
\

w
W

w
K

a
1.5

I U

w 1.0 w

w z w
0
0

z
.5

vapor pressure of A temperature dependence, O R . vapor pressure constant for A vapor pressure of B temperature dependence, O R . c 4 vapor pressure constant for B specific heat condenser water, B.t.u. (1b.-OR.)- 1 c w E = activaiton energy, B.t.u./lb.-mole F = flow rate, lb./hr. Fw = water flow rate, Ib./hr. H = holdup, lb. AH, = heat of vaporization, B.t.u./lb. A H , = heat of reaction, B.t.u./lb. H , = holdup of water in condenser, lb. Kc = feedback controller gain, lb. (hr.-OR.) - 1 K E Q = mass transfer coefficient, Ib. (hr.-lb./sq.in.) - 1 k = specific reaction rate, hr.-l M = molecular weight, lb. (1b.-mole) -1 N = total number of stages P = pressure, lb./sq.in.abs. Pa = vapor pressure A, lb./sq.in.abs. Ps = vapor pressure B, lb./sq.in.abs. P, = condenser Dressure. lb./sa.in.abs. = reactor pressure, Ib./sq.infabs. = overall reaction rate, hr.-l = perfect gas constant, B.t.u. (1b.-mole O R . ) = vapor flow resistance between reactor and condenser, lb./sq.in. (lb./hr.) - 2 = ratio molecular weights = rativ of vapor pressures = C1 == C3, O R . = temperature, O R . = time, hours = condenser process temperature, OR. = reactor vapor temperature, O R . = condenser water temperature, "R. = inlet water temperature, OR. = condenser heat transfer coefficient, B.t.u. (hr.sq.ft.-OR.) -1 = vapor rate, lb./hr. = interstage mixing rate, Ib./hr. = mole fraction A in reactor liquid = mole fraction A in condenser liquid = mole fraction A in vapor = weight fraction A in vapor = pre-exponential factor, hr.-1
= = = = =

LITERATURE CITED
0

.2

.4

.6
(HOURS)

.8

1.0

1. Luyben, W. L., AIChE j . , 12,662 (1966). 2. Van Heerden, C., Ind. Eng. Chem., 45, 1242 (1953).
Manuscript receitied September 1 I , 1967; revision receitied November 30, 1967; paper accepted December 4 , 1967.

TIME

Fig. 7. Batch autorefrigerated reactor.

Vol. 14, No. 6

AlChE Journal

Page 885